Satellite-based Aerosol Optical Depth Research Articles

Montana Statewide Google Earth Engine-Based Wildfire Hazardous Particulate (PM2.5) Concentration Estimation

Wildfires pose a direct threat to the property, life, and well-being of the population of Montana, USA, and indirectly to their health through hazardous smoke and gases emitted into the atmosphere. Studies have shown that elevated levels of particulate matter cause impacts to human health ranging from early death, to neurological and immune diseases, to cancer. Although there is currently a network of ground-based air quality sensors (n = 20) in Montana, the geographically sparse network has large gaps and lacks the ability to make accurate predictions for air quality in many areas of the state. Using the random forest method, a predictive model was developed in the Google Earth Engine (GEE) environment to estimate PM2.5 concentrations using satellite-based aerosol optical depth (AOD), dewpoint temperature (DPT), relative humidity (RH), wind speed (WIND), wind direction (WDIR), pressure (PRES), and planetary-boundary-layer height (PBLH). The validity of the prediction model was evaluated using 10-fold cross validation with a R2 value of 0.572 and RMSE of 9.98 µg/m3. The corresponding R2 and RMSE values for ‘held-out data’ were 0.487 and 10.53 µg/m3. Using the validated prediction model, daily PM2.5 concentration maps (1 km-resolution) were estimated from 2012 to 2023 for the state of Montana. These concentration maps are accessible via an application developed using GEE. The product provides valuable insights into spatiotemporal trends of PM2.5 concentrations, which will be useful for communities to take appropriate mitigation strategies and minimize hazardous PM2.5 exposure.

Evaluation and comparison of MODIS and MISR aerosol products with ground-based monitoring stations in the Amazon Basin

This study evaluates the long-term accuracy of six daily aerosol optical depth (AOD) datasets derived from the Moderate Resolution Imaging Spectroradiometers (MODIS) Dark Target (DT) at 3 and 10 km resolution, Deep Blue (DB), combined DT and DB datasets (DTB) and MODIS Multi-Angle Implementation of Atmospheric Correction (MAIAC) algorithms and the Multi-angle Imaging SpectroRadiometer (MISR) Version 23, by comparing them with ground-based AOD observations in 18 Aerosol Robotic Network (AERONET) sites over the Amazon basin between 2000 to 2022. The AOD retrieval accuracy was also assessed under varying seasons, aerosol loading, particle size, elevation and land cover type. Overall results showed that Terra and Aqua-MODIS DB and MISR algorithms perform better, with about 88% of AOD retrievals within the expected error range. However, MISR presented a systematic high negative bias. MODIS DT 3km overestimated AOD values and indicated unsatisfactory retrieval accuracy in almost all scenarios. MODIS and MISR retrievals showed improved performance under low aerosol loading, whereas Aqua-MODIS MAIAC under intermediate conditions (AOD>0.2). MISR indicated a large dependence on a coarse mode-dominated environment with considerable negative bias. Terra-MODIS MAIAC retrievals exhibited a smaller dependence for aerosol particle size distribution, while Terra-MODIS DB performed better for fine mode and Aqua-MODIS DB for coarse dominance. MAIAC also performed well under polluted scenarios in the dry season in contrast to the majority of satellite-based AOD retrievals, which showed good agreement under clean background conditions in the wet season. Almost all satellite-based retrievals showed higher uncertainty under extremely high altitudes (>3000m) followed by non-satisfactory correlation and significant biases. Additionally, almost all satellite-based AOD retrievals indicated poor performance as vegetation coverage increases. On the contrary, MISR showed higher accuracy under forest type. In general, MODIS and MISR retrieval accuracy demonstrated distinct levels of dependencies that require further improvement in the Amazon Basin.

First retrieval of daily 160 m aerosol optical depth over urban areas using Gaofen-1/6 synergistic observations: Algorithm development and validation

The satellite-based aerosol optical depth (AOD), which can provide continuous spatial observations of aerosol loadings, is widely adopted to estimate atmospheric environmental quality and evaluate its risk for human health. However, current satellite-retrieved AOD products characterized by a comparatively coarse spatial resolution (≥1 km) can hardly analyze the structure of atmospheric pollution or its correlation with urban landscapes over populated urban areas. Existed studies have tried to address this deficiency by retrieving high-resolution AOD using Landsat images, whose long revisit period (16 days nominally), however, largely limits its applications related to urban air pollution research. To achieve both high spatial resolution and expected revisit period from satellite observation, in this study, a comprehensive AOD retrieval framework is developed for Wide-Field-of-View (WFV) satellite sensors to yield a daily AOD dataset over urban areas with 160 m spatial resolution, based on the Gaofen-1 and Gaofen-6 synergistic observations. To address the crucial challenge that the high spatial resolution and complex urban landscape both contribute a dramatic variation of land surface bidirectional reflectance, a Simulated-Annealing-coupled Semiempirical Modified Rahman-Pinty-Verstraete (SAS-MRPV) scheme is proposed to model and estimate the land surface reflectance (LSR) in the AOD retrieval framework, where the SAS-MRPV scheme is implemented using simulated annealing iteration initialized by the bidirectional reflectance distribution function (BRDF) products from the Moderate Resolution Imaging Spectrometer (MODIS) as a priori knowledge. The validation results demonstrate that the Gaofen WFV AOD retrievals exhibit good agreement with the ground-based AErosol RObotic NETwork (AERONET) and SONET (Sun-sky radiometer Observation NETwork) AOD, demonstrating the correlation coefficient and the expected error (EE)±(0.05 + 0.15AODAERONET/SONET) ratio respectively reaching up to 0.97 and over 80 %, and only slight bias fluctuations are found under different land cover types, aerosol loading, and seasonal conditions. Moreover, the validation results of Gaofen WFV AOD retrievals, with LSR estimated based on the SAS-MRPV scheme, the MODIS BRDF Products, and the Minimum Reflectance Technique (MRT) method, demonstrate better accuracy and completeness of AOD retrievals using the SAS-MPRV scheme over both natural and impervious land surfaces, indicating the stability and reliability of proposed SAS-MRPV LSR determination scheme in WFV AOD retrieval. In addition to that, the error analyses and quality control results demonstrate that the SAS-MRPV scheme is effective and necessary to guarantee the reliability and accuracy of land surface bidirectional reflectance estimation by identifying and excluding the land surface pixels unsuitable for LSR modeling, which mitigates the uncertainty and yield better accuracy in the final WFV AOD products. Furthermore, the inter-comparison between WFV-derived AOD against the operational MODIS Deep Blue (10 km), Dark Target (3 km), and Multi-Angle Implementation of Atmospheric Correction (MAIAC, 1 km) AOD products demonstrates that the 160 m WFV AOD retrievals, on the basis of obtaining similar aerosol overall descriptions as MODIS AOD retrievals, possess the capability to characterize the spatial distribution of atmospheric pollution at a finer scale with smoother variation under both clean and polluted conditions. With outstanding accuracy and reliable performance, the developed comprehensive high-resolution AOD retrieval framework exhibits substantial potential for operational AOD retrieval of Gaofen WFV satellite observations, which could be directly applied to other urban areas supporting further studies related to air pollution emission management and health risk assessment at extra-fine spatial scale.

Effect of COVID-19 Lockdown on Urban Heat Island Dynamics in Prague, Czechia

Urban heat islands (UHI) are a well-known phenomenon adversely affecting human health and urban environments. The worldwide COVID-19 lockdown in 2020 provided a unique opportunity to investigate the effects of decreased emission of air pollution and anthropogenic heat flux (AHF) on UHI. Although studies have suggested that reduced AHF during lockdown decreased atmospheric UHI (AUHI) and surface UHI (SUHI), these results contain inherent uncertainties due to unaccounted weather variability and urban-rural dynamics. Our study comprehensively analyzes the impact of the COVID-19 lockdown on AUHI and SUHI in Prague, Czechia. By selecting days with similar weather conditions, we examined changes in mean SUHI using MODIS satellite images and in AUHI based on air temperature from Prague weather stations for the Lockdown period during March–April 2020 versus a Reference period from March–April 2017–2019. Our results show that, in comparison to the Reference period, the Lockdown period was associated with a 15% (0.1 °C) reduction of SUHI in urbanized areas of Prague and a 0.7 °C decline in AUHI in the city center. Additionally, the observed decreases in satellite-based aerosol optical depth and nitrogen dioxide by 12% and 29%, respectively, support our hypothesis that the weakened UHI effects were linked to reduction in anthropogenic activities during the lockdown. Revealing the largest decrease of mean SUHI magnitude around the periphery of Prague, which has predominantly rural land cover, our study emphasizes the need to consider the effects of urban-rural dynamics when attributing changes in SUHI to AHF. Our findings provide additional insights into the role of reduced anthropogenic activities in UHI dynamics during the COVID-19 lockdown and offer policymakers a comprehensive understanding of how the complex interaction between urban and rural microclimate dynamics influences the SUHI phenomenon.

A deep learning-based imputation method for missing gaps in satellite aerosol products by fusing numerical model data

Satellite-based aerosol optical depth (AOD) products are commonly used in various aerosol-related studies, such as aerosol pollution mapping and aerosol-climate interactions. However, these satellite AOD products often suffer from significant missing gaps due to cloud cover and limitations in the retrieval algorithm. To address this issue, some studies take advantage of real-time seamless simulation of numerical models and successfully fill in these gaps by establishing a regression relationship between satellite AOD and numerical model AOD. However, these previous studies usually use satellite AOD retrievals as the regression target, which limits the accuracy of the imputation results by the original accuracy of satellite AOD retrievals and also consumes a considerable amount of time. To overcome these limitations, this study proposes a spatiotemporal imputation model called Bi-ConvRNN, which combines convolutional neural networks (CNN) and bidirectional recurrent neural networks (Bi-RNN). The model takes both satellite AOD retrievals and numerical model AOD data as input and utilizes the weighted mean squared error (MSE) loss function of multiple AOD datasets, e.g., ground-based data, satellite retrievals, and numerical simulation, as the optimization target to improve the imputation accuracy. The proposed model is evaluated using hourly COMS GOCI AOD products. In the independent test set, the AOD results generated by the Bi-ConvRNN model in the region containing GOCI AOD retrievals can break the accuracy of original GOCI AOD products with the accuracy improved from R2 = 0.70 [RMSE = 0.15] to R2 = 0.84 [RMSE = 0.11], and the filling accuracy, e.g. R2 = 0.79, [RMSE = 0.14], in the region without GOCI AOD retrievals are still better than those of the original GOCI AOD retrievals. Additionally, the Bi-ConvRNN model demonstrates satisfactory filling efficiency, requiring only 0.12 s to fill in the missing gaps of hourly GOCI AOD products per day. These results highlight the efficiency and reliability of the proposed model in filling the gaps in satellite AOD products, and the filled AOD results have great potential for further aerosol-related research.

Satellite-based aerosol optical depth estimates over the continental U.S. during the 2020 wildfire season: Roles of smoke and land cover

Wildfires produce smoke that can affect an area >1000 times the burn extent, with far-reaching human health, ecologic, and economic impacts. Accurately estimating aerosol load within smoke plumes is therefore crucial for understanding and mitigating these impacts. We evaluated the effectiveness of the latest Collection 6.1 MODIS Multi-Angle Implementation of Atmospheric Correction (MAIAC) algorithm in estimating aerosol optical depth (AOD) across the U.S. during the historic 2020 wildfire season. We compared satellite-based MAIAC AOD to ground-based AERONET AOD measurements during no-, light-, medium-, and heavy-smoke conditions identified using the Hazard Mapping System Fire and Smoke Product. This smoke product consists of maximum extent smoke polygons digitized by analysts using visible band imagery and classified according to smoke density. We also examined the strength of the correlations between satellite- and ground-based AOD for major land cover types under various smoke density levels. MAIAC performed well in estimating AOD during smoke-affected conditions. Correlations between MAIAC and AERONET AOD were strong for medium- (r = 0.91) and heavy-smoke (r = 0.90) density, and MAIAC estimates of AOD showed little bias relative to ground-based AERONET measurements (normalized mean bias = 3 % for medium, 5 % for heavy smoke). During two high AOD, heavy smoke episodes, MAIAC underestimated ground-based AERONET AOD under mixed aerosol (i.e., smoke and dust; median bias = −0.08) and overestimated AOD under smoke-dominated (median bias = 0.02) aerosol. MAIAC most overestimated ground-based AERONET AOD over barren land (mean NMB = 48 %). Our findings indicate that MODIS MAIAC can provide robust estimates of AOD as smoke density increases in coming years. Increased frequency of mixed aerosol and expansion of developed land could affect the performance of the MAIAC algorithm in the future, however, with implications for evaluating wildfire-associated health and welfare effects and air quality standards.

Integrating Augmented In Situ Measurements and a Spatiotemporal Machine Learning Model To Back Extrapolate Historical Particulate Matter Pollution over the United Kingdom: 1980-2019.

Historical PM2.5 data are essential for assessing the health effects of air pollution exposure across the life course or early life. However, a lack of high-quality data sources, such as satellite-based aerosol optical depth before 2000, has resulted in a gap in spatiotemporally resolved PM2.5 data for historical periods. Taking the United Kingdom as an example, we leveraged the light gradient boosting model to capture the spatiotemporal association between PM2.5 concentrations and multi-source geospatial predictors. Augmented PM2.5 from PM10 measurements expanded the spatiotemporal representativeness of the ground measurements. Observations before and after 2009 were used to train and test the models, respectively. Our model showed fair prediction accuracy from 2010 to 2019 [the ranges of coefficients of determination (R2) for the grid-based cross-validation are 0.71-0.85] and commendable back extrapolation performance from 1998 to 2009 (the ranges of R2 for the independent external testing are 0.32-0.65) at the daily level. The pollution episodes in the 1980s and pollution levels in the 1990s were also reproduced by our model. The 4-decade PM2.5 estimates demonstrated that most regions in England witnessed significant downward trends in PM2.5 pollution. The methods developed in this study are generalizable to other data-rich regions for historical air pollution exposure assessment.

Retrieval of nighttime aerosol optical depth by simultaneous consideration of artificial and natural light sources

Aerosol Optical Depth (AOD) is a critical optical parameter that quantifies the degree of light attenuation by aerosols and serves as a fundamental indicator of atmospheric quality. Therefore, accurate quantification and retrieval of AOD is crucial for relevant studies. However, current satellite-based AOD retrieval algorithms suffer from inapplicability under low-light conditions, limiting the development of nighttime AOD retrieval. Under this context, we proposed a novel algorithm, namely Simultaneous Consideration of Artificial and Natural light Sources (SCANS), to obtain nighttime AOD. The core of the SCANS algorithm is considering the synergy of both the natural and artificial light sources to obtain nighttime AOD by integrating atmospheric radiative transfer simulation into an extinction method and performing multiple iterations. SCANS was applied to the Visible Infrared Imaging Radiometer Suite (VIIRS) Day/Night Band (DNB) and the retrieved nighttime AOD was validated with in-situ measurements from five AERONET sites. Results indicate that the Mean Bias Errors (MBEs) of the retrieved nighttime AOD range from 0.0 to 0.08 and the corresponding Root Mean Square Errors (RMSEs) range from 0.11 to 0.17, which exhibit better accuracy than that of the nighttime MERRA-2 AOD. We further compared the retrieved nighttime AOD with the corresponding Air Quality Index (AQI) measurements at six environment monitoring stations and obtained high correlation coefficients (i.e., ranging from 0.733 to 0.940), indicating SCANS's reliability and high accuracy. The proposed SCANS algorithm can effectively obtain nighttime AOD with high quality, thereby advancing research on the diurnal variation of crucial Earth's key elements.

Satellite-based study of physico-optical properties of aerosols over a westernmost location of Brahmaputra valley

This study examines the long term trend of the radiatively active atmospheric aerosols which can influence the Earth’s energy budget directly by scattering and absorbing radiation and indirectly by acting as cloud condensation nuclei. MODIS sensor on board the NASA Earth Observing System Terra and Aqua satellite based Aerosol Optical Depth (AOD) data are used for long term analysis of aerosols over Bongaigaon, Assam for the period August, 2002 to March, 2017. Highest AOD values are observed in pre-monsoon (March-May) season due to long range transportation as well as intense biomass burning activities especially as a part of Jhum cultivation. In general, AOD values are low in post-monsoon (October-November) season which may be due to wash out of aerosols by rain in the preceding months without enough replacement. The monthly AOD values vary from its highest value 0.949 in April, 2016 to its lowest value 0.107 in November, 2002 for the study period. From the comparison of MODIS Terra and Aqua AOD at 550 nm, it is clearly seen that generally Terra AOD at 10:30 hr is higher than the Aqua AOD at 13:30hr. A slowly increasing trend of both Aqua and Terra AOD at 550 nm is observed over the study location. The observed Ångström exponent value varies from its minimum value in monsoon season to its maximum value in winter season. With increasing AOD values, horizontal visibility decreases over Bongaigaon.

Spectral and Spatial Dependencies in the Validation of Satellite-Based Aerosol Optical Depth from the Geostationary Ocean Color Imager Using the Aerosol Robotic Network

Spectral and Spatial Dependencies in the Validation of Satellite-Based Aerosol Optical Depth from the Geostationary Ocean Color Imager Using the Aerosol Robotic Network

Generation of High Temporal Resolution Full-Coverage Aerosol Optical Depth Based on Remote Sensing and Reanalysis Data

Aerosol Optical Depth (AOD) is a crucial physical parameter used to measure the radiative and scattering properties of the atmosphere. Obtaining full-coverage AOD measurements is essential for a thorough understanding of its impact on climate and air quality. However, satellite-based AOD products can be affected by abnormal weather conditions and high reflectance surfaces, leading to gaps in spatial coverage. To address this issue, we propose a satellite-based AOD filling method based on hourly level-3 Himawari-8 AOD products. In this study, the optimal model with a mean bias error (MBE) less than 0.01 and a root-mean-square error (RMSE) less than 0.1 in most land cover types was selected to generate the full-coverage AOD. The generated full-coverage AOD was validated against in situ measurements from the AERONET sites and compared with the performance of Himawari-8 AOD and MERRA-2 AOD over the AERONET sites. The validation results indicate that the accuracy of full-coverage AOD is comparable to that of the Advanced Himawari Imager (AHI) AOD, and for other land cover types (excluding barren land), the accuracy of full-coverage AOD is superior to that of MERRA-2 AOD. To investigate the practical application of full-coverage AOD, we utilized it as an input parameter to perform radiative transfer simulations in northwestern and southern China. The validation results showed that the simulated at-sensor radiance based on full-coverage AOD was in good agreement with the at-sensor radiance observations from MODIS. These results indicate that complete and accurate measurements of AOD have considerable potential for application in the simulation of at-sensor radiance and other related topics.

Window-Based Filtering Aerosol Retrieval Algorithm of Fine-Scale Remote Sensing Images: A Case Using Sentinel-2 Data in Beijing Region

The satellite-based Aerosol Optical Depth (AOD) retrieval algorithms are generally needed to construct Land Surface Reflectance (LSR) database. However, errors are unavoidable due to the surface complexity, especially for the short observation period and high-resolution images, such as Sentinel-2 Multi-Spectral Instrument (MSI) data. To address this, reference day images are used instead of the LSR database. The surface is assumed to be Lambertian; however, the fact is that not all pixels meet it well. Therefore, we proposed a window-based AOD retrieval algorithm, which can ignore the unreliable/non-Lambertian pixels in a retrieval window based on two main filtering processes. Finally, using Sentinel-2 Band 1 (60 m), the AODs (120 m) of 134 reference images to 43 reference images were retrieved by this algorithm from 2017 to 2021 in Beijing region, China. The results show that the retrieved AOD with the proposed algorithm exhibits good agreement with the ground-based measured AOD (R > 0.97). The high-resolution AOD presents comparable spatial distributions to the Multiangle Implementation of Atmospheric Correction (MAIAC) algorithm AOD (1 km) products. Moreover, the very little noise and very high spatial continuity of retrieval AOD imply that this algorithm could be ported to other algorithms as part of improving AOD quality.

How aerosol size matters in aerosol optical depth (AOD) assimilation and the optimization using the Ångström exponent

Abstract. Satellite-based aerosol optical depth (AOD) has gained popularity as a powerful data source for calibrating aerosol models and correcting model errors through data assimilation. However, simulated airborne particle mass concentrations are not directly comparable to satellite-based AODs. For this, an AOD operator needs to be developed that can convert the simulated mass concentrations into model AODs. The AOD operator is most sensitive to the input of the particle size and chemical composition of aerosols. Furthermore, assumptions regarding particle size vary significantly amongst model AOD operators. More importantly, satellite retrieval algorithms rely on different size assumptions. Consequently, the differences between the simulations and observations do not always reflect the actual difference in aerosol amount. In this study, the sensitivity of the AOD operator to aerosol properties has been explored. We conclude that, to avoid inconsistencies between the AOD operator and retrieved properties, a common understanding of the particle size is required. Accordingly, we designed a hybrid assimilation methodology (hybrid AOD assimilation) that includes two sequentially conducted procedures. First, aerosol size in the model operator has been brought closer to the assumption of the satellite retrieval algorithm via assimilation of Ångström exponents. This ensures that the model AOD operator is more consistent with the AOD retrieval. The second step in the methodology concerns optimization of aerosol mass concentrations through direct assimilation of AOD (standard AOD assimilation). The hybrid assimilation method is tested over the European domain using Moderate Resolution Imaging Spectroradiometer (MODIS) Deep Blue products. The corrections made to the model aerosol size information are validated through a comparison with the ground-based Aerosol Robotic Network (AERONET) optical product. The increments in surface aerosol mass concentration that occur due to either the standard AOD assimilation analysis or the hybrid AOD assimilation analysis are evaluated against independent ground PM2.5 observations. The standard analysis always results in relatively accurate posterior AOD distributions; however, the corrections are hardly transferred into better aerosol mass concentrations due to the uncertainty in the AOD operator. In contrast, the model AOD and mass concentration states are considerably more accurate when using the hybrid methodology.

Aerosol optical depth regime over megacities of the world

Abstract. Currently, 55 % of the world's population resides in urban areas and this number is projected to increase to 70 % by 2050. Urban agglomerations with a population over 10 million, characterized as megacities, are expected to be more than 100 by 2100. Such large concentrations of population could boost creativity and economic progress, but also raises several environmental challenges such as air quality degradation. In this study, we investigate the spatial and temporal variability of urban aerosol state of 81 cities with a population over 5 million, relying on daily satellite-based aerosol optical depth (AOD) retrievals, derived at fine spatial resolution (0.1∘×0.1∘), over an 18-year period spanning from 2003 to 2020. According to our results, the lowest long-term mean AOD values worldwide were found in European and American cities (from 0.08 to 0.20). For almost all African and Asian cities, mean AOD ranged from 0.25 up to 0.90, but a considerable dust aerosol contribution (up to 70 %) was found for some of them with associated mean dust optical depth (DOD) values reaching up to 0.4. Mostly Chinese and Indian cities tend to have higher mean AOD values in the areas surrounding their center, while the opposite was found for most of the cities in the rest of the world. High intraannual AOD variability was revealed for the eastern American cities, while lower values were found in Chinese, eastern Indian and the eastern Mediterranean cities. During the study period, statistically significant negative AOD decadal trends were found for East Asian, European and North American cities, with the greatest decrease of −0.1 to −0.3 per decade recorded for the Chinese cities, in which the maximum mean AODs (0.45–0.91) are observed. In most of the US cities, where low mean AOD <0.17 was recorded, considerable declining AOD trends were found (−30 % to −50 % per decade). For the rest of Asian, African and South American cities, statistically significant AOD increase was found, with the greatest values of +0.07 to +0.16 per decade recorded for Indian cities. In Bengaluru (India), it is reported the lowest mean AOD value (0.2) and the maximum AOD increase (+69 %), which may be partially attributed to the population growth over the study period. The agreement of the satellite-derived AOD trends against those obtained from ground-based AERONET measurements was examined. For ground-based stations within the geographical limits of the contiguous urban area of the examined cities, a 0.93 correlation for the long-term means of AOD was found and ∼75 % of the derived trends agreed in sign. It was found that the spatial homogeneity within the examined satellite domain and the location of the surface station were key factors that determined their agreement. The present study highlights the vital and essential contribution of spaceborne products to monitor aerosol burden over megacities of the planet towards fulfilling the United Nations Sustainable Development Goal of “sustainable cities and communities”, dealing with urban air quality.

Prediction of daily mean and one-hour maximum PM2.5 concentrations and applications in Central Mexico using satellite-based machine-learning models.

Machine-learning algorithms are becoming popular techniques to predict ambient air PM2.5 concentrations at high spatial resolutions (1 × 1 km) using satellite-based aerosol optical depth (AOD). Most machine-learning models have aimed to predict 24 h-averaged PM2.5 concentrations (mean PM2.5) in high-income regions. Over Mexico, none have been developed to predict subdaily peak levels, such as the maximum daily 1-h concentration (max PM2.5). Our goal was to develop a machine-learning model to predict mean PM2.5 and max PM2.5 concentrations in the Mexico City Metropolitan Area from 2004 through 2019. We present a new modeling approach based on extreme gradient boosting (XGBoost) and inverse-distance weighting that uses AOD, meteorology, and land-use variables. We also investigated applications of our mean PM2.5 predictions that can aid local authorities in air-quality management and public-health surveillance, such as the co-occurrence of high PM2.5 and heat, compliance with local air-quality standards, and the relationship of PM2.5 exposure with social marginalization. Our models for mean and max PM2.5 exhibited good performance, with overall cross-validated mean absolute errors (MAE) of 3.68 and 9.20 μg/m3, respectively, compared to mean absolute deviations from the median (MAD) of 8.55 and 15.64 μg/m3. In 2010, everybody in the study region was exposed to unhealthy levels of PM2.5. Hotter days had greater PM2.5 concentrations. Finally, we found similar exposure to PM2.5 across levels of social marginalization. Machine learning algorithms can be used to predict highly spatiotemporally resolved PM2.5 concentrations even in regions with sparse monitoring. Our PM2.5 predictions can aid local authorities in air-quality management and public-health surveillance, and they can advance epidemiological research in Central Mexico with state-of-the-art exposure assessment methods.

Lidar-based daytime boundary layer height variation and impact on the regional satellite-based PM2.5 estimate

Satellite-derived aerosol optical depth (AOD) is popularly used to infer ground-level PM2.5 concentration due to its wide coverage. The fact that aerosols are largely confined in the atmospheric boundary layer makes boundary layer height (BLH) an important scale factor for AOD-based PM2.5 estimates. Our recent ground-based lidar observations, nevertheless, indicate that aerosol particles are heterogeneously mixed within the boundary layer, and even frequently reside above BLH, forming the residual layers (RL)-like pattern. To better sort out the underlying mechanism behind the above-mentioned phenomenon and the impact on hourly ground-level PM2.5 estimates from satellite-based AOD, we firstly propose a novel notion of haze layer height (HLH), which is calculated from Micro-Pulse Lidar (MPL) profile. Combined analysis of 3.5-year ground-based lidar profiles, CE-318 AOD, and PM2.5 measurements show that the coefficient of determination (R2) between PM2.5 and AOD normalized with HLH increases from 0.49 to 0.61 for 90% of the dataset. Second, we applied HLH to an Auto-encoder-based Deep Residual Network (ADRN) and tested the effect on satellite AOD-based PM2.5 estimation within a 300 km range surrounding the MPL station. With the aid of the AOD imputation technique, a similar improvement of using HLH instead is found on the regional scale PM2.5 estimation, which can be demonstrated by the comparison with air quality measurements and other machine learning models. The results show that using ADRN with HLH achieves the highest performance (mean R2 = 0.87, RMSE = 10.12 μg/m3) among 4 machine learning models. This new approach, largely combining active and passive remote sensing data through artificial neural networks (CAPTA), shows improved accuracy and coverage of hourly PM2.5 estimation with aerosol vertical information and AOD calculation under clouds. In addition, further analysis showed that the average difference between morning and daily PM2.5 concentration could equate to an accuracy of 0.19–2.57 yrs. in terms of life expectancy, indicating that our new approach to the determination of PM2.5 from space sheds new insight into the assessment of the aerosol impact on public health.

Bayesian Aerosol Retrieval-Based PM2.5 Estimation through Hierarchical Gaussian Process Models

Satellite-based aerosol optical depth (AOD) data are widely used to estimate land surface PM2.5 concentrations in areas not covered by ground PM2.5 monitoring stations. However, AOD data obtained from satellites are typically at coarse spatial resolutions, limiting their applications on small or medium scales. In this paper, we propose a new two-step approach to estimate 1-km-resolution PM2.5 concentrations in Shanghai using high spatial resolution AOD retrievals from MODIS. In the first step, AOD data are refined to a 1×1km2 resolution via a Bayesian AOD retrieval method. In the second step, a hierarchical Gaussian process model is used to estimate PM2.5 concentrations. We evaluate our approach by model fitting and out-of-sample cross-validation. Our results show that the proposed approach enjoys accurate predictive performance in estimating PM2.5 concentrations.

Using satellite-based AOD and ground-based measurements to evaluate the impact of the DWH oil spill on coastal air quality

The 2010 DWH disaster generated atmospheric pollutants of health concern which reached the Gulf Coast. This study evaluated whether changes in coastal air quality due to the disaster were captured by aerosol optical depth (AOD) estimated using satellite data and by ground-based monitoring of air pollution, including fine particulate matter ≤2.5 μm in aerodynamic diameter (PM2.5), benzene and naphthalene. Mean monthly AOD levels were higher in May 2010 [during oil spill time], (mean AOD = 0.355), than for the prior (mean AOD = 0.258) and following years (mean AOD = 0.252) (p < 0.05). PM2.5 concentrations and AOD were significantly correlated (R2 = 0.59, p < 0.05), for one study area. Elevated PM2.5, benzene, and naphthalene concentrations coincided with downwind directions from the location of the oil slicks. A fully-coupled oil fate and transport atmospheric transport model of oil spill emissions, integrated with AOD and more extensive ground-based measurements, is recommended to predict coastal population exposures during oil spills.

Studies on the improvement of modelled solar radiation and the attenuation effect of aerosol using the WRF-Solar model with satellite-based AOD data over north China

Photovoltaic (PV) power generation will become one of the main energy sources to achieve goals of carbon peak and carbon-neutral in China in the future. However, serious haze events have occurred frequently due to massive anthropogenic emissions in autumn and winter in China and resulted in a significant reduction of solar radiation reaching the Earth's surface. In this study, we conducted modifications to modelled solar radiation with the Weather Research and Forecasting (WRF)-Solar model by implementing the satellite-based aerosol optical depth (AOD) data into the model. The satellite-based AOD data have a higher spatial resolution and better accuracy. The studies were performed during the heavy haze period in the winter of 2016/2017 over North China (NC). Comparison with observations shows that the temporal-spatial variations of satellite-based AOD are consistent with those of AERONET and surface PM2.5 concentration, and satellite-based AOD at different sites or some months is slightly underestimated. Satellite-based AOD and surface PM2.5 concentrations in most of the southern areas of NC are significantly larger than in the northern parts. Simulated errors of Global horizontal irradiance (GHI) are reduced remarkably compared with those excluding the AOD data, especially in regions of the BTH (Beijing, Tianjin, and Hebei province), west of Shandong province, and north of Henan province. All RMSE and MAE of simulated GHI under three types of weather conditions declined by more than 20%, and 37%, respectively. Modelled errors of GHI on all-sky and cloudy-sky days are larger than those on clear-sky days due to the complicated aerosol-cloud-radiation interaction and simulated bias of cloud. Reductions of GHI caused by aerosols fall in between -140～-20 W m−2, and the spatial distribution is more refined and perform well compared to observations. Our results show that direct application of the satellite-based AOD data into the WRF-Solar model is very effective for improving the simulation of solar radiation and investigating the attenuation effect of solar radiation caused by aerosol in winter.

Himawari-8/AHI Aerosol Optical Depth Detection Based on Machine Learning Algorithm

Due to the advantage of geostationary satellites, Himawari-8/AHI can provide near-real-time air quality monitoring over China with a high temporal resolution. Satellite-based aerosol optical depth (AOD) retrieval over land is a challenge because of the large surface contribution to the top of atmosphere (TOA) signal and the uncertainty of aerosol modes. Here, by combining satellite TOA reflectance, sun-sensor geometries, meteorological factors and vegetation information, we propose a data-driven AOD detection algorithm based on a deep neural network (DNN) model for Himawari-8/AHI. It is trained by sample data of 2018 and 2019 and is applied to derive hourly AODs over China in 2020. By comparison with ground-based AERONET measurements, R2 for DNN-estimated AOD is up to 0.8702, which is much higher than that for the AHI AOD product with R2 = 0.4869. The hourly AOD results indicate that the DNN model has a good potential in improving the performance of AOD retrieval in the early morning and in the late afternoon, and the spatial distribution is reliable for capturing the variation of aerosol pollution on the regional scale. By analyzing different DNN modeling strategies, it is found that seasonal modeling can hardly increase the accuracy of AOD retrieval to a certain extent, and R2 increases from 0.7394 to 0.8168 when meteorological features, especially air pressure, are involved in the model training.